Forecasting Risk of Crop Disease with Anomaly Detection Algorithms

Information from crop disease surveillance programs and outbreak investigations provides real-world data about the drivers of epidemics. In many cases, however, only information on outbreaks is collected surrounding healthy crops are omitted. Use such to develop models that can forecast risk/no risk therefore problematic, as relating no-risk status missing. This study explored a novel application anomaly detection techniques derive for forecasting composed only. was done in two steps. training phase, algorithms were used learn envelope weather conditions most associated with historic outbreaks. testing hindcasting events. Five different compared according their accuracy outbreaks: robust covariance, one-class k-means, Gaussian mixture model, kernel density estimation, support vector machine. A case potato late blight survey across Great Britain proof concept. The results showed model had highest at 97.0%, followed by k-means 96.9%. There added value combining an ensemble provide more accurate tool be tailored produce region-specific alerts. here easily applied other pathosystems tools agricultural decision support.